Fooling LLM graders into giving better grades through neural activity guided adversarial prompting

Open in new window